Digital Humanities |
Computational Social Science
As a fun but relevant side project to my research, I thought I'd start a series about a week of Twitter discussion. The much-encompassing, similar yet different topics: Ursula von der Leyen and Angela Merkel.
My research deals with Opinion Dynamics and persuasion, in particular how different communities might consider different forms of argumentation and information to be persuasive. It is a pretty labor-intensive field of inquiry where you see results late in the game (but you can speed things up). So, I decided to get my "hands dirty" and try out different approaches as a side project. I also do this series, because...well, visualization is fun.
I look at two political figures that have been discussed in the last week: Angela Merkel and Ursula von der Leyen. Why these two? They are both high-profile politicians that are controversially discussed for different reasons, many relevant political topics are discussed in association with their name and, most importantly, they were simply available as topics with a high tweet volume when I started data collection. Angela Merkel in particular had some bouts of physical weakness in public and Ursula von der Leyen has been elected President of the European Commission, which sparked somewhat polarized discussions.
A lot of generally important topics are discussed in association with these two public figures. The discussions also reflect the day-to-day political activity on German Twitter (moreso for the discussion on Merkel). While of course important, big events like the European Parliament elections show the best and worst of the Twitter population, thoroughly looking at the more day-to-day political activity could deliver some very interesting information about how it operates more generally.
I will first give a coarse birds-eye view about what happened on Twitter (hashtag networks, general activity and user networks). In the next few weeks, I will then dive in to the data and show in more detail how the topics and arguments are distributed over time, how different communities differ in activity, what sources they use and, conclusively, all of this taken together. As you may know, it's important to know "when did who say what to whom, and through which channel" to understand a discourse more deeply. Only knowing which hashtags were important overall does not tell you that much. I guess this will make about four or five large posts, but I might decide to go for shorter, more frequent ones (my dayjob requires some time, also, after all).
In particular, follower/following networks and replies are hard to get from Twitter, so my data is somewhat incomplete in that regard. If you want to help me out with data collection, you can donate a Twitter Access Token (see Help Me Out. If you're interested in how I did this, you can take a look at the documentation on my GitHub (link will follow here in the next few days).
Now, without further ado, let's look at the data!
We start with the discussion on Ursula von der Leyen. For the timeframe of the 7th of July up until the 17th, we see 89.597 tweets by 34.848 users. You can see right away that there were two significant events on the 10th, 11th of July as well as on the 16th, 17th and 18th.
You can already guess that the second spike comes directly after the news of her election. The first spike is the announcement of the election date. Discussion is somewhat scaled back after the announcement and revs up when the result hits the news.
You could argue that most of the action happens in the replies and you can see that here: a clear spike in reply activity. The thing that sparks the most is of course the Retweets, as it is easy and meant for information dissemination. The second spike phase almost has as much original Tweets as Retweets, unlike the first spike, which was more about spreading the initial information.
As I said, I probably do not have all of the replies, so there is some margin of error here (shameless third cry for help).
Now, let's look at the discussion on Angela Merkel. The overall volume is comparable. We have 94.388 tweets by 33150 users. However, the dynamics are quite different.
This is somewhat different: Lower volume overall (look at the y axis), but it is constant. Think about it: people are always talking about Angela Merkel, she is the most visible politician in Germany. So even though there are certain events that stand out (e.g., bouts of physical weakness), she is a more constant topic of day-to-day discussion. The spikes have to do with Ursula von der Leyens nomination, which of course also concerns Merkel, and Angela Merkels birthday (17th of July), which I coincidentally only found out about as I first looked at the data.
While the overall volume is quite comparable, the dynamics over time differ. As I will show down below, the producers of these tweet and those that are often RT'd, replied to and mentioned, also differ somewhat, although we see similar topics.
Different kinds of topics with different activity patterns is already a topic in research. Angela Merkel is a constant topic with only low fluctuations, but Ursula von der Leyen (this past week) has been more like an event - short spikes of more extreme activity.
But what kind of activity?
What looks like the Death Star below is the hashtag network for the discussion on von der Leyen. It is somewhat mixed and I had a hard time in getting it into a readably visualization. Take care when looking at this: where the hashtags are in the picture does not reflect actual distance, only to some degree. I ran an algorithm over the data that further assigns the same colors to hashtags that occur together very often. But this is not scientific fact, only a help when deciding what to look at next. The sizes of the nodes reflect the "importance" they have within the network.
Hashtag network for the discussion on Ursula von der Leyen (click for a large, zoomable PDF version)
Even though it is largely mixed, we can see some things that form coherent topics. The light blue color seems to be about von der Leyen, her previous role as Minister of Defense and the CDU itself (Annegret Kramp-Karrenbauer, akk, and Jens Spahn are clustered together).
The light green and light pink colors indicate the political process of the EU itself ("spitzenkandidat", "wahl", "bruessel", etc.), while most of the important, big nodes are concerned with the EU and von der Leyens new role. We also have a big node "demokratie" (democracy) which is tied to many other hashtags - the election process for the president of the EU commission is often deemed undemocratic.
To the left of the large vonderleyen node we find a network of critical hashtags: flintenuschi ("shotgun ursula"), berateraffäre, incompetenz (incompetence), gorchfock, zensursula. Flintenuschi is simply a demeaning nickname, much like other hashtags that use the fact the "leyen" sounds like another word for "amateur". The berateraffäre is a scandal of the Minsitry of Defense concerning seemingly unnecessary grants for consulting firms and the Gorch Fock is a ship that should have been restored, but mismanagement by von der Leyens department lead to unforeseeable cost. "Zensursula" is a nickname that plays with von der Leyens opinion that "problematic" content on the web should be censored (thus "zensur"), which here is connected to protests against the recent EU copyright reforms (article13). The "piraten" party (pirate party) is also within this cluster, so we can assume that these more critical voices often come from supporters.
Hashtag network for the discussion on Ursula von der Leyen (click for a large, zoomable PDF version)
On the top right we also see some conspiracy and racist theories. Here we find mentions of black people ("neger"), something about jews ("juden") and "genealogy" as well as "puppets". Some of this seems to tightly overlap with the Merkel hashtag network (i.e., the same users were active here).
I will go into more detail in future posts, but feel free to look for interesting connections yourself.
The Merkel hashtag network is more clear-cut:
We see the events in the news reflected here: Angela Merkel has publicly lauded Greta Thunberg for her climate activism, which we can see right at the top ("fridaysforfuture", "co2steuer"(C02 tax)) and further to the right, in the middle ("greta"). Her public trembling is also adressed (e.g., "zitteranfall", "merkelzittern") and if you squint, you can find some congratulations to her birthday.
We also see the Seawatch and the actions of Carola Rackete discussed, which is a generally relevant topic in Germany and Europe itself.
However, how this large chunk came to be, I have no idea. Like in the von der Leyen network, we see presumably see some racism ("neger", "juden") as well as Seawatch, Trump and his Huawei deal ("lifeline") among various other things. What this important cluster is supposed to be a more detailed analysis will have to show.
I plan to show which clusters of users are more active in certain topics in future posts. But which users, anyway?
I constructed these user networks via the retweets, mentions and replies. Getting the actual follower / following relations takes way too much time (again, you can help me out here), but these networks reflect the actual activity better, anyway. You can see pretty well which users have the most incoming communications (bigger). I excluded labels for users that are seldomly tweeted at, because if they're not in the public light they shouldn't be dragged into it. The same color means that the users are interacting more in some way or other.
The relations here are pretty obvious. To the left, we kind of see a satirical front: Martin Sonneborn, the heuteshow (satirical news show), Nico Semsrott and DiePartei. Semsrott and Sonneborn are satirists associated with DiePartei ("the party") that have been elected into the European Parliament (now in the European Green Party). Semsrott also had aspirations to be elected President of the European Commission, which might explain why these users are so prominent.
In the middle, you will find politicians of the German Left (Sarah Wagenknecht), SPD (Sven Giegold) as well as the Green Party (Tiemo Wölken, Ska Keller).
In the upper right, you can find the accounts of politicians of the German far-right and euroskeptic party Alternative für Deutschland (red color).
The user network for the discussion on Merkel draws a different picture:
The satirists are much less important here, or at least it's not the same satirists. This might be because the more European topic von der Leyen was more important. However, much more interesting is the fact that the far-right bubble is composed differently.
Instead of only the official accounts of politicians, we now find normal user accounts to be much more important in the right sphere. We also see Hans-Georg Maaßen in a very prominent role: the former president of the Federal Office for the Protection of the Constitution (Bundesamt für Verfassungsschutz) has previously supported the far-right AfD under different circumstances. However, if his importance here is for the same reason remains to be shown.
Interestingly, we also see some activity from the Austrian Twittersphere. Sebastian Kurz is important in this network and we also see many journalists and activists from Austria in the orange blob to the left.
How these users interact with different topics, I will try to show in future posts. There seems to be a different audiences for the discussion of Ursula von der Leyen and Angela Merkel. Merkel is more reflective of general topics in Germany.
Next time, I will look at the differences between retweet, reply and mention networks. Feel free to write me with suggestions and criticism under @Boomcrashkapow or under firstname.lastname@example.org.